Date: March 18, 2013
Author(s): Robert Tanner
Looking for a mainstream SSD but need help deciding which to choose? Is that $10 difference really worth it? Who’s got the most reliable SandForce based-SSD around? Why does Intel have so many SandForce SSDs anyway? All these questions and more are answered within!
Intel should need no introduction when it comes to solid-state drives – after all, it is generally recognized that the company’s entry into the SSD market back in 2008 with the X25 series is what catalyzed the SSD movement. The market has changed considerably since then, and similarly so has Intel’s SSD offerings.
Today, Intel’s consumer lineup is comprised entirely of SandForce SF-2281 powered drives. Even so, the company’s reputation as one of the most reliable SSD options in the market has not changed, thanks to a full year of validation on its 520 family before launch. It didn’t just write its own custom firmware for the SandForce controller, either – even the hardware design was tweaked in its quest to purge errata.
The 330 and 335 Series are similar, with the main differences being NAND die size (The 520 gets the best binned NAND, hence the 5 year warranty and price premium to match). Interestingly, one unexpected surprise we noticed was with the power consumption; Intel is claiming to have cut both idle and load power draw by more than half of the 330/520 models – an impressive feat thanks to IMFT’s smaller, more efficient 20nm MLC NAND.
|Intel Consumer SSDs|
|SSD 330 (180GB)||SSD 335 (180GB)||SSD 520 (180GB)|
|Capacities (GB)||60, 120, 180, 240||180, 240||60, 120, 180, 240, 480|
|NAND||Intel 25nm MLC||Intel 20nm MLC||Intel 25nm MLC|
|Sequential Read||500 MB/s||550MB/s|
|Sequential Write||450 MB/s||520 MB/s|
|Max Random Reads||42,000 IOPS||50,000 IOPS|
|Max Random Writes||52,000 IOPS||80,000 IOPS|
|Power (Idle/Load)||600mW / 850mW||275mW / 350mW||600mW / 850mW|
|Warranty||3 Years||5 Years|
The above table represents the current listing of Intel’s SSD products, with the sole exception the 325 Series which uses a different, mSATA connector. Notably, the 335 ships solely with a 9.5mm height form factor, making it a notable exception to the recent trend of 7mm drives. Users looking for 7mm height Intel drives will have to resort to the 520 or 525 Series.
The graphic on the front of the drive is still notably absent, although Intel indicates at some future point 335’s will be featuring the case redesign. The back of the housing also sports the Intel-typical unfinished manufacturing look that we’ve seen before on the 520. Thankfully, most of these drives will be looked at once and then squirreled away to not be seen again for a few years.
Package contents include the following: quick-start guide, metal 2.5″-to-3.5″ plate adapter, speed demon sticker, both sets of mounting screws, molex-to-SATA power adapter and a SATA cable. The latest version of Intel’s Toolbox utility can be downloaded from here and here for the in-Windows cloning utility.
Intel’s Toolbox utility offers pretty much every function the user would want for using the SSD. It provides an estimated life remaining count that shows how much of the expected program / erase cycles have been used, although even then most NAND will have additional cycles well after the 0% mark has been reached. Users can secure erase the SSD, update firmware, check SMART data and even run a sector level check of the entire drive.
At Techgage, we strive to make sure our results are as accurate and real-world applicable as possible. We list most of the steps and processes involved in setting up and conducting our benchmarking process below, but in the interests of brevity we can’t mention every last detail. If there is any pertinent information that we’ve inadvertently omitted or you have any thoughts, suggestions, or critiques, then please feel free to email us or post directly in our forums. This site exists for readers like you and we value your input.
The table below lists the hardware used in our current storage-testing machine, which remains unchanged throughout all of our testing, with the obvious exception of the storage device. Each drive used for the sake of comparison is also listed here.
|Techgage Solid-State Drive Test System|
|Processor||Intel Core i7-2600 – 3.80GHz (Locked) Quad-Core|
|Motherboard||ASUS P8P67 Deluxe|
|Memory||4GB Kingston DDR3-2133|
|Graphics||AMD Radeon HD 5770|
|Storage||Hitachi 7200RPM 2TB Hard Drive|
Corsair Neutron GTX 240GB
Crucial m4 256GB
Intel 335 Series 180GB
Kingston HyperX 3K 240GB
Kingston SSDNow V300 240GB
OCZ Vector 256GB
OCZ Vertex 4 256GB
|Power Supply||Antec NeoHE 550W|
|Et cetera||Dell 2407WFP (1920×1200)|
Windows 7 Ultimate SP1 64-bit
Our Windows 7 Desktop for SSD Testing
When preparing our SSD testbed for benchmarking we follow these guidelines:
Windows 7 Optimizations
For our new Sandy Bridge storage testbed we have migrated to using test images for our drives. All drives are imaged with the cloned test image to ensure all drivers, programs, and settings remain identical for testing purposes. We feel disk cloning software and SSD controller technology has matured to the point where potential issues such as non-aligned sectors are no longer a potential issue.
For testing, we run all tests five times dropping the highest and lowest results, then take the average of the middle three. And who said that college statistics class wouldn’t prove useful? If any anomalous results are seen the test will be run again. Given the complexities of modern computers, and especially today’s operating systems and the software that runs on them, we feel this provides the most accurate results possible.
Finally, we are seeking to constantly improve and expand upon our SSD testing methodology. We are always actively seeking real-world workload scenarios that are bottlenecked by hard drives, so if you have any suggestions whatsoever or there is a program you would like to see included in our SSD content, then please drop by our forums and let us know! We are always looking to expand our SSD benchmarks and provide more useful and real-world results, and not just synthetic numbers.
Futuremark’s PCMark benchmarking suite should need no introduction; it has been a staple of PC benchmarks for the better half of a decade. It includes over 25 individual workloads designed to measure all aspects of system performance and gives individual scores in each test as well as an overall system performance score for easy system comparisons.
PCMark 7 offers a more accurate measure of performance as compared to its predecessor, PCMark Vantage. The storage scoring metrics especially were significantly re-tuned and optimized with SSDs in mind to give a more balanced disk subsystem score.
Although IMFT’s 20nm NAND is rated for a slightly lower write speed than the older 25nm NAND, it doesn’t tangibly impact performance. The Intel 335 still outperforms the m4 in PCMark and places between the two other SSDs in our lineup that utilize SandForce controllers. Surprisingly, the 335 is even able to swap places with the Neutron in the storage composite score.
Originally developed by Intel – and since given to the open-source community – Iometer (pronounced “eyeawmeter”, like thermometer) is one of the best storage-testing applications available, for a couple of reasons. The first, and primary, is that it’s completely customizable, and if you have a specific workload you need to test a drive with, you can easily accomplish it here. Secondly, it bypasses the Windows disk subsystem entirely, meaning it bypasses the OS drivers and writes directly to the storage media. This has important implications, such as it means Windows 7 cannot correctly align Iometer to match the SSD or HDD sector alignment.
We have updated our test suite to the latest stable 1.10 rc1 build of Iometer, which was released in December, 2010. This version makes some changes to be aware of; specifically, it gives the option for three types of data sets used during testing. 2006 and earlier versions used a pseudo-random dataset for testing, while the 1.10 build will default to a “repeating bytes” test pattern. A full random test mode was also added. To avoid giving SandForce drives an unfair advantage (they rely on data compression to achieve their performance), we will stick to the pseudo-random test pattern for all of our testing.
We have configured Iometer for correct 4KB disk alignment using a single 8GB test file from within Windows, meaning they are acting as the host OS drive with no other drives in the system. We run individual random 4KB read and write tests at a queue depth of 3 and again at 32. Then we run the 128KB sequential read & write tests using a queue depth of 1. In addition, all drives are in a dirty state prior to testing – this means results will not be comparable to advertised manufacturer results. Our goal is to measure end-user performance under real-world conditions, and so our testing reflects typical SSD performance after it has been used for some length of time in a system. Each test pattern is run for 5 minutes to achieve an average result.
In addition, we have created three Iometer disk usage scenarios that should roughly approximate database, file server, and workstation usage patterns. These scenarios are run individually for 10 minutes each within an 8GB file on the drive, which is an unusually harsh scenario for any sort of SSD. Drives that are able to offer better sustained performance over time and those that favor certain file size accesses will do well here. All three tests are configured for a queue depth of 32 to show which drives are best capable of dealing with heavy workload scenarios.
“IOPS” is simply the measure of performance relative to a certain disk access size, specifically 4KB or 512 bytes, or any size desired. Typically with SSDs when speaking about IOPS it is referred to on the assumption of 4KB accesses. With this in mind, it is easy to convert between IOPS and MB/s. Iometer provides both types of results to us and for the sake of concise graphs, brevity, and easily understandable results, we have elected to use MB/s for the 4KB and 128KB tests. For reference: IOPS = (MBps Throughput / KB per IO) * 1024 and MBps = (IOPS * KB per IO) / 1024.
In the read portion of the tests the 335 performs well, again slotting directly between both other SandForce drives, although right behind the leader as opposed to the SandForce value drive.
In the write tests things get a bit more interesting, with Intel’s drive easily outperforming both other SandForce drives. The m4 may offer better sequential performance, but for average consumer use, random 4KB performance is more important – in that respect the 335 easily trounces the m4.
As for the three usage scenarios, the Intel 335 performs well and stays in the middle of the pack with no dips or performance issues. The Intel 335 can’t compete at the high-end SSD market, but it wasn’t ever intended to do so.
Overall, the 335 180GB does well, offering well-rounded performance without any single point of weakness. Clearly, Intel has optimized the 335 firmware with the best balance between sequential reads & writes and random reads & writes without compromising any one of the four areas for the sake of better performance in the other three.
As the name implies, AS SSD is a nifty little program written exclusively for solid-state drives. It can still be run on a mechanical hard drive just for fun, but be warned: what takes a few minutes on an SSD will require the better part of an hour on an HDD! It is freely available for download here.
This handy tool measures sequential reads and writes in addition to the important 4KB random reads and writes, then ranks the results with a final score for quick comparison with other SSDs. In addition to the main test there is a secondary benchmark that simulates the type of data transferred for ISO, Program, and Game files. We selected this program for its precision, ability to generate large file sizes on-the-fly, and because it is written to bypass Windows 7’s automatic caching system.
Second only to Iometer, AS SSD is one of the best tools for distinguishing between multiple SSDs while providing a scoring system for easier, quick comparisons that anyone can run for themselves.
In the read tests, the 335 does fine sequentially in the 4KB test. But in typical Intel fashion, at higher queue depths 4KB read performance climbs significantly and is able to easily slip ahead of the m4. We say “typically” because Intel has always optimized its firmware to perform best at a queue depth of 32.
In the write tests the 335 doesn’t offer any surprises, but does deliver solid performance across all three data patterns. It also does well in the three file tests.
The final score is based primarily on the four corners of SSD performance (sequential reads & writes, and random reads & writes) and is split into individual read and write scores as well as a final overall score, with the Intel drive able to deliver solid, even levels of performance.
HD Tune is still primarily an HDD benchmark, but we include it as an alternative for those consumers that prefer it for one reason or another. The free version does not perform write tests, but otherwise is available for free here.
HD Tune has always produced differing results compared to other synthetic programs, and this time is no different. There is clearly something about the Intel 335 that HD Tune likes, as it does very well in the sequential read test and performs second only to the Vector in the random access tests. Just as with AS SSD, access latency times are also favorable.
Finally, we reach the first of our real-world tests where there are no unusual testing or scoring algorithms to leave us scratching our heads, just simple tests to see how an SSD changes actual system performance.
For the File Transfer test we took a 4.5GB compressed archive and measured how much time was required to transfer the file to another folder on the same drive. Keep in mind that with a hard disk, this requires the actuator arm to seek back and forth between the source and destination sectors on the disk platter, with the destination sectors often not sequentially aligned. In contrast, any SSD can concurrently perform read and write operations simultaneously on any NAND chip without regard to spatial considerations of bits strewn randomly around a disk platter, which gives them a large advantage here.
The Intel 335 didn’t favor sequential writes in our previous tests, but it performs decently here with average file copy performance.
Either you’ve heard of FLAC, or it is an integral part of your digital life. But iTunes and Apple devices do not support FLAC files, leaving those with discerning ears forced to use Apple’s Lossless codec. dBpoweramp makes it possible to convert between them utilizing as many threads as are available to the system.
In this test, we take 10 albums amounting to 4GB of FLAC files and convert them to Apple’s lossless format. This creates exactly 3.96GB of new data. This scenario is even more applicable for those users with six or more physical CPU cores available, because as the core count increases, the more the storage system will become the actual bottleneck. Our test rig is limited to only a quad-core processor, but even then we can see clear differences amongst the various contenders.
No unexpected surprises here, the 335 does well and again slots in the middle of the pack and in the middle of both the other SandForce drives. It also again manages to edge out the m4 by a comfortable six second margin.
Real-world results are surprisingly hard to come by when testing SSDs. It is extremely easy to showcase just how much faster any SSD on the market is compared to even a modern mechanical disk drive. However, when we try to compare SSD to SSD, differences can amount to just a few seconds or even a fraction of a second, often well inside the margin of error (and human reflexes), making any results obtained meaningless.
We are always eager to hear about any demanding storage workloads our readers may have, but in an effort to get around this problem, we have put together three batch test files that target three levels of intensity.
Firstly we have our light batch file, which we drop into the Windows Startup folder. Windows 7 will execute and load various programs and commands as it boots, making it perhaps the most easily pertinent of our three tests. Almost everyone has an array of programs that starts with their OS, ranging from background applications like anti-virus to programs like a browser or music player.
This batch file will load four websites in Firefox, start Photoshop CS5 and load five 5MB or greater images, and load 15MB of data in Word, Excel, and Powerpoint documents. Several background utilities will also load; a PDF file and compressed file are opened for viewing, and of course, since nobody likes to work without listening to some music, we have our favorite 56MB FLAC file playing the entire time. Obviously, all of this takes place while Windows 7 itself is still loading. We start timing from the moment the machine is powered on to the moment the last program finishes loading – and it isn’t as long as you might think. (We provide raw cold boot times on the next page for direct comparison).
Our medium batch test is similar although we apply the use of timers to space apart the commands. Instead of booting, time begins from the moment we execute the batch file until the moment all tasks have completed. The medium test also consists of the following:
To keep things simple, the heavy batch test is identical to the medium test in all respects, save for one important addition. Computer users coming from HDDs will be familiar with the slowdown or even molasses-like feeling that occurs from having an anti-virus or anti-malware scan running in the background. SSDs scoff at this sort of thing however, and the typical SSD user wouldn’t think twice about running an anti-virus scan at the same time as playing a fullscreen game since framerates will remain relatively unaffected.
The heavy test will capitalize on this by running an anti-virus scan from Microsoft Security Essentials on a static, unchanging 5.1GB test folder that contains 19,748 files and 2,414 sub-folders copied from the Program Files directory. Also worth noting is that because the medium and heavy batch tests are identical save for the AV scan, results between them are directly comparable.
Light batch results are mixed but the 335 does well here. For the medium batch tests it slips behind and ties the m4 with 93 seconds. But once we throw in the AV scan, the higher queue depth optimizations that Intel favors kick in and let it shave 15 seconds off the m4’s heavy batch time. Overall, the 335’s heavy scenario result compares favorably for a mainstream drive and slots in significantly better than the also-SandForce powered V300.
For the boot test, we perform a cold boot with the stopwatch starting the moment the power button is pressed until the last systray icon has finished loading. A large number of factors can change how fast a computer starts; whether the motherboard uses a BIOS or the newer UEFI; if a RAID controller has to be initialized; to delay timers or other motherboard optimizations. In other words, individual results will vary depending on the system hardware.
Obviously, our P67 motherboard is the bottleneck at this point; it takes more than just fast storage to quick-boot a system! Regardless, there are no unexpected results to see here.
SSDs deliver some of the most benefits to games. Not only can the game load significantly faster so users can hurry up and wait to get through various advertisement screens, but they also boost level or map load times. For games where player immersion into the new world is important, the difference between 15 and 25 seconds can seem huge when waiting for the next part of the level or world to load.
For our new regimen we chose Portal 2 and Civilization V. Portal 2 is already a very well optimized game, but it’s immersive, so we time how long it takes to load the sp_a2.bts6 custscene. With Civilization V‘s recent overhaul to game storage files to help decrease load times, and the new option to disable the intro movie trailer, it becomes possible to time how long it takes to start the game.
Game level load times are good as well, although it’s safe to say any SSD will deliver significantly better map load and level transition times than a mechanical drive.
As expected for an Intel solid-state drive, the 335 performs admirably without any shortcomings. It certainly isn’t the fastest SSD on the block, but again, it was never intended to be as it is aimed squarely at the mainstream consumer market. The 335 still performs best with compressible data, as underneath the hood, it’s still a SandForce SF-2281 controller. But even so, Intel’s firmware team has put together a well-rounded option that favors high queue-depth workloads.
Intel has always held a small price premium on its SSD products, and the 335 Series is no exception. That said, current pricing on the 335 family is still unnaturally high due to the still very low supply of these drives. This is why the 330 SSDs can usually be found for less, and in this instance we have no problem recommending users looking to buy an Intel 335 to also consider the 330 Series. As the drives themselves are identical (save for different NAND), performance will be nearly identical between them, and at least for the time being users can save some money by sticking to the 330 family. The only exception to this may be laptop or ultra-mobile users that want every last minute of battery life, in which case the 335 would likely be worth an extra couple bucks just for its large cut in power consumption.
Intel’s reputation for reliability is well-earned, but there are also concrete reasons for it. Just like with processors, some NAND comes out of the oven better than others, and Intel’s joint partnership with Micron gives Intel first dibs on all NAND that IMFT fabricates. This means Intel traditionally gets to choose from the higher binned NAND for its products. The 20nm 8GB MLC NAND used in the 335 Series is still new, but reportedly employs the same high-k/metal gate structure as its processors and will match the same 3,000 program / erase cycles as the larger 25nm NAND found in the 330 Series.
But access to better NAND isn’t all. Intel spent a considerable amount of time creating its own proprietary, custom firmware for the SF-2281 controller from scratch, as well as baking in its own proven methods for minimizing write amplification and improving wear leveling (which it pioneered with its first in-house SSDs). This has gone a considerable way toward insulating Intel from the litany of problems the SF-2281 controller faced after its initial reception and kept the company’s first SandForce drive, the 520 Series, in high regard. Whether this is worth a slight price premium over competing SSDs, we shall leave entirely up to each consumer buying them – but against competing SandForce drives it is definitely an option to keep in mind.
Copyright © 2005-2019 Techgage Networks Inc. - All Rights Reserved.